Blog
Technical notes from the prokube team about Kubernetes, MLOps, LLMs, agentic AI, and the infrastructure patterns behind production AI systems.
Latest article
A complete KServe setup for Qwen3.8 27B, including a reusable vLLM runtime, BF16 and FP8 deployment settings, H100 NVL validation, and the most important tuning knobs.
Why "gateway" in cloud native and AI now means almost anything, and how to avoid losing the plot completely.
How we ran the Gemma 4 31B instruction-tuned model on a single A100 80GB GPU using KServe and vLLM, including the custom runtime, PVC setup, tuning flags, verification, monitoring, and troubleshooting notes.